1 /* FLEX lexer for Ada expressions, for GDB. -*- c++ -*-
2 Copyright (C) 1994-2022 Free Software Foundation, Inc.
4 This file is part of GDB.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 3 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 /*----------------------------------------------------------------------*/
21 /* The converted version of this file is to be included in ada-exp.y, */
22 /* the Ada parser for gdb. The function yylex obtains characters from */
23 /* the global pointer lexptr. It returns a syntactic category for */
24 /* each successive token and places a semantic value into yylval */
25 /* (ada-lval), defined by the parser. */
28 NUM10 ({DIG}({DIG}|_)*)
30 NUM16 ({HEXDIG}({HEXDIG}|_)*)
33 ID ({LETTER}({LETTER}|{DIG}|[\x80-\xff])*|"<"{LETTER}({LETTER}|{DIG})*">")
36 GRAPHIC [a-z0-9 #&'()*+,-./:;<>=_|!$%?@\[\]\\^`{}~]
37 OPER ([-+*/=<>&]|"<="|">="|"**"|"/="|"and"|"or"|"xor"|"not"|"mod"|"rem"|"abs")
44 #include "diagnostics.h"
46 /* Some old versions of flex generate code that uses the "register" keyword,
47 which clang warns about. This was observed for example with flex 2.5.35,
48 as shipped with macOS 10.12. The same happens with flex 2.5.37 and g++ 11
49 which defaults to ISO C++17, that does not allow register storage class
52 DIAGNOSTIC_IGNORE_DEPRECATED_REGISTER
54 #define NUMERAL_WIDTH 256
55 #define LONGEST_SIGN ((ULONGEST) 1 << (sizeof(LONGEST) * HOST_CHAR_BIT - 1))
57 /* Temporary staging for numeric literals. */
58 static char numbuf[NUMERAL_WIDTH];
59 static void canonicalizeNumeral (char *s1, const char *);
60 static struct stoken processString (const char*, int);
61 static int processInt (struct parser_state *, const char *, const char *,
63 static int processReal (struct parser_state *, const char *);
64 static struct stoken processId (const char *, int);
65 static int processAttribute (const char *);
66 static int find_dot_all (const char *);
67 static void rewind_to_char (int);
70 #define YY_DECL static int yylex ( void )
72 /* Flex generates a static function "input" which is not used.
73 Defining YY_NO_INPUT comments it out. */
77 #define YY_INPUT(BUF, RESULT, MAX_SIZE) \
78 if ( *pstate->lexptr == '\000' ) \
82 *(BUF) = *pstate->lexptr; \
84 pstate->lexptr += 1; \
87 static int find_dot_all (const char *);
89 /* Depth of parentheses. */
90 static int paren_depth;
94 %option case-insensitive interactive nodefault noyywrap
102 "--".* { yyterminate(); }
105 canonicalizeNumeral (numbuf, yytext);
106 char *e_ptr = strrchr (numbuf, 'e');
108 return processInt (pstate, nullptr, numbuf, e_ptr + 1);
112 canonicalizeNumeral (numbuf, yytext);
113 return processInt (pstate, NULL, numbuf, NULL);
116 {NUM10}"#"{HEXDIG}({HEXDIG}|_)*"#"{POSEXP} {
117 canonicalizeNumeral (numbuf, yytext);
118 char *e_ptr = strrchr (numbuf, 'e');
120 return processInt (pstate, numbuf,
121 strchr (numbuf, '#') + 1,
125 /* The "llf" is a gdb extension to allow a floating-point
126 constant to be written in some other base. The
127 floating-point number is formed by reinterpreting the
128 bytes, allowing direct control over the bits. */
129 {NUM10}(l{0,2}f)?"#"{HEXDIG}({HEXDIG}|_)*"#" {
130 canonicalizeNumeral (numbuf, yytext);
131 return processInt (pstate, numbuf, strchr (numbuf, '#') + 1,
136 canonicalizeNumeral (numbuf, yytext+2);
137 return processInt (pstate, "16#", numbuf, NULL);
141 {NUM10}"."{NUM10}{EXP} {
142 canonicalizeNumeral (numbuf, yytext);
143 return processReal (pstate, numbuf);
147 canonicalizeNumeral (numbuf, yytext);
148 return processReal (pstate, numbuf);
151 {NUM10}"#"{NUM16}"."{NUM16}"#"{EXP} {
152 error (_("Based real literals not implemented yet."));
155 {NUM10}"#"{NUM16}"."{NUM16}"#" {
156 error (_("Based real literals not implemented yet."));
159 <INITIAL>"'"({GRAPHIC}|\")"'" {
160 yylval.typed_val.val = yytext[1];
161 yylval.typed_val.type = type_for_char (pstate, yytext[1]);
165 <INITIAL>"'[\""{HEXDIG}{2,}"\"]'" {
166 ULONGEST v = strtoulst (yytext+3, nullptr, 16);
167 yylval.typed_val.val = v;
168 yylval.typed_val.type = type_for_char (pstate, v);
172 /* Note that we don't handle bracket sequences of more than 2
173 digits here. Currently there's no support for wide or
174 wide-wide strings. */
175 \"({GRAPHIC}|"[\""({HEXDIG}{2,}|\")"\"]")*\" {
176 yylval.sval = processString (yytext+1, yyleng-2);
181 error (_("ill-formed or non-terminated string literal"));
186 rewind_to_char ('i');
191 rewind_to_char ('t');
195 thread{WHITE}+{DIG} {
196 /* This keyword signals the end of the expression and
197 will be processed separately. */
198 rewind_to_char ('t');
205 and { return _AND_; }
206 else { return ELSE; }
211 null { return NULL_PTR; }
213 others { return OTHERS; }
215 then { return THEN; }
218 /* BOOLEAN "KEYWORDS" */
220 /* True and False are not keywords in Ada, but rather enumeration constants.
221 However, the boolean type is no longer represented as an enum, so True
222 and False are no longer defined in symbol tables. We compromise by
223 making them keywords (when bare). */
225 true { return TRUEKEYWORD; }
226 false { return FALSEKEYWORD; }
230 {TICK}[a-z][a-z_]+ { BEGIN INITIAL; return processAttribute (yytext+1); }
234 "=>" { return ARROW; }
235 ".." { return DOTDOT; }
236 "**" { return STARSTAR; }
237 ":=" { return ASSIGN; }
238 "/=" { return NOTEQUAL; }
242 <BEFORE_QUAL_QUOTE>"'" { BEGIN INITIAL; return '\''; }
244 [-&*+./:<>=|;\[\]] { return yytext[0]; }
246 "," { if (paren_depth == 0 && pstate->comma_terminates)
248 rewind_to_char (',');
255 "(" { paren_depth += 1; return '('; }
256 ")" { if (paren_depth == 0)
258 rewind_to_char (')');
268 "."{WHITE}*all { return DOT_ALL; }
271 yylval.sval = processId (yytext+1, yyleng-1);
275 {ID}({WHITE}*"."{WHITE}*({ID}|\"{OPER}\"))*(" "*"'")? {
276 int all_posn = find_dot_all (yytext);
278 if (all_posn == -1 && yytext[yyleng-1] == '\'')
280 BEGIN BEFORE_QUAL_QUOTE;
283 else if (all_posn >= 0)
285 yylval.sval = processId (yytext, yyleng);
290 /* GDB EXPRESSION CONSTRUCTS */
292 "'"[^']+"'"{WHITE}*:: {
294 yylval.sval = processId (yytext, yyleng);
298 "::" { return COLONCOLON; }
300 [{}@] { return yytext[0]; }
302 /* REGISTERS AND GDB CONVENIENCE VARIABLES */
304 "$"({LETTER}|{DIG}|"$")* {
305 yylval.sval.ptr = yytext;
306 yylval.sval.length = yyleng;
307 return DOLLAR_VARIABLE;
310 /* CATCH-ALL ERROR CASE */
312 . { error (_("Invalid character '%s' in expression."), yytext); }
316 /* Initialize the lexer for processing new expression. */
319 lexer_init (FILE *inp)
327 /* Copy S2 to S1, removing all underscores, and downcasing all letters. */
330 canonicalizeNumeral (char *s1, const char *s2)
332 for (; *s2 != '\000'; s2 += 1)
343 /* Interprets the prefix of NUM that consists of digits of the given BASE
344 as an integer of that BASE, with the string EXP as an exponent.
345 Puts value in yylval, and returns INT, if the string is valid. Causes
346 an error if the number is improperly formated. BASE, if NULL, defaults
347 to "10", and EXP to "1". The EXP does not contain a leading 'e' or 'E'.
351 processInt (struct parser_state *par_state, const char *base0,
352 const char *num0, const char *exp0)
356 /* For the based literal with an "f" prefix, we'll return a
357 floating-point number. This counts the the number of "l"s seen,
358 to decide the width of the floating-point number to return. -1
360 int floating_point_l_count = -1;
367 base = strtol (base0, &end_of_base, 10);
368 if (base < 2 || base > 16)
369 error (_("Invalid base: %d."), base);
370 while (*end_of_base == 'l')
372 ++floating_point_l_count;
375 /* This assertion is ensured by the pattern. */
376 gdb_assert (floating_point_l_count == -1 || *end_of_base == 'f');
377 if (*end_of_base == 'f')
380 ++floating_point_l_count;
382 /* This assertion is ensured by the pattern. */
383 gdb_assert (*end_of_base == '#');
389 exp = strtol(exp0, (char **) NULL, 10);
392 while (isxdigit (*num0))
394 int dig = fromhex (*num0);
396 error (_("Invalid digit `%c' in based literal"), *num0);
397 mpz_mul_ui (result.val, result.val, base);
398 mpz_add_ui (result.val, result.val, dig);
404 mpz_mul_ui (result.val, result.val, base);
408 if (floating_point_l_count > -1)
410 struct type *fp_type;
411 if (floating_point_l_count == 0)
412 fp_type = language_lookup_primitive_type (par_state->language (),
413 par_state->gdbarch (),
415 else if (floating_point_l_count == 1)
416 fp_type = language_lookup_primitive_type (par_state->language (),
417 par_state->gdbarch (),
421 /* This assertion is ensured by the pattern. */
422 gdb_assert (floating_point_l_count == 2);
423 fp_type = language_lookup_primitive_type (par_state->language (),
424 par_state->gdbarch (),
428 yylval.typed_val_float.type = fp_type;
429 result.write (gdb::make_array_view (yylval.typed_val_float.val,
430 TYPE_LENGTH (fp_type)),
431 type_byte_order (fp_type),
437 gdb_mpz maxval (ULONGEST_MAX / base);
438 if (mpz_cmp (result.val, maxval.val) > 0)
439 error (_("Integer literal out of range"));
441 LONGEST value = result.as_integer<LONGEST> ();
442 if ((value >> (gdbarch_int_bit (par_state->gdbarch ())-1)) == 0)
443 yylval.typed_val.type = type_int (par_state);
444 else if ((value >> (gdbarch_long_bit (par_state->gdbarch ())-1)) == 0)
445 yylval.typed_val.type = type_long (par_state);
446 else if (((value >> (gdbarch_long_bit (par_state->gdbarch ())-1)) >> 1) == 0)
448 /* We have a number representable as an unsigned integer quantity.
449 For consistency with the C treatment, we will treat it as an
450 anonymous modular (unsigned) quantity. Alas, the types are such
451 that we need to store .val as a signed quantity. Sorry
452 for the mess, but C doesn't officially guarantee that a simple
453 assignment does the trick (no, it doesn't; read the reference manual).
455 yylval.typed_val.type
456 = builtin_type (par_state->gdbarch ())->builtin_unsigned_long;
457 if (value & LONGEST_SIGN)
458 yylval.typed_val.val =
459 (LONGEST) (value & ~LONGEST_SIGN)
460 - (LONGEST_SIGN>>1) - (LONGEST_SIGN>>1);
462 yylval.typed_val.val = (LONGEST) value;
466 yylval.typed_val.type = type_long_long (par_state);
468 yylval.typed_val.val = value;
473 processReal (struct parser_state *par_state, const char *num0)
475 yylval.typed_val_float.type = type_long_double (par_state);
477 bool parsed = parse_float (num0, strlen (num0),
478 yylval.typed_val_float.type,
479 yylval.typed_val_float.val);
485 /* Store a canonicalized version of NAME0[0..LEN-1] in yylval.ssym. The
486 resulting string is valid until the next call to ada_parse. If
487 NAME0 contains the substring "___", it is assumed to be already
488 encoded and the resulting name is equal to it. Similarly, if the name
489 starts with '<', it is copied verbatim. Otherwise, it differs
491 + Characters between '...' are transfered verbatim to yylval.ssym.
492 + Trailing "'" characters in quoted sequences are removed (a leading quote is
493 preserved to indicate that the name is not to be GNAT-encoded).
494 + Unquoted whitespace is removed.
495 + Unquoted alphabetic characters are mapped to lower case.
496 Result is returned as a struct stoken, but for convenience, the string
497 is also null-terminated. Result string valid until the next call of
501 processId (const char *name0, int len)
503 char *name = (char *) obstack_alloc (&temp_parse_space, len + 11);
505 struct stoken result;
508 while (len > 0 && isspace (name0[len-1]))
511 if (name0[0] == '<' || strstr (name0, "___") != NULL)
513 strncpy (name, name0, len);
522 if (isalnum (name0[i0]))
524 name[i] = tolower (name0[i0]);
527 else switch (name0[i0])
542 while (i0 < len && name0[i0] != '\'');
553 /* Return TEXT[0..LEN-1], a string literal without surrounding quotes,
554 with special hex character notations replaced with characters.
555 Result valid until the next call to ada_parse. */
558 processString (const char *text, int len)
562 const char *lim = text + len;
563 struct stoken result;
565 q = (char *) obstack_alloc (&temp_parse_space, len);
570 if (p[0] == '[' && p[1] == '"' && p+2 < lim)
572 if (p[2] == '"') /* "...["""]... */
580 ULONGEST chr = strtoulst (p + 2, &end, 16);
582 error (_("wide strings are not yet supported"));
592 result.length = q - result.ptr;
596 /* Returns the position within STR of the '.' in a
597 '.{WHITE}*all' component of a dotted name, or -1 if there is none.
598 Note: we actually don't need this routine, since 'all' can never be an
599 Ada identifier. Thus, looking up foo.all or foo.all.x as a name
600 must fail, and will eventually be interpreted as (foo).all or
601 (foo).all.x. However, this does avoid an extraneous lookup. */
604 find_dot_all (const char *str)
608 for (i = 0; str[i] != '\000'; i++)
615 while (isspace (str[i]));
617 if (strncasecmp (str + i, "all", 3) == 0
618 && !isalnum (str[i + 3]) && str[i + 3] != '_')
624 /* Returns non-zero iff string SUBSEQ matches a subsequence of STR, ignoring
628 subseqMatch (const char *subseq, const char *str)
630 if (subseq[0] == '\0')
632 else if (str[0] == '\0')
634 else if (tolower (subseq[0]) == tolower (str[0]))
635 return subseqMatch (subseq+1, str+1) || subseqMatch (subseq, str+1);
637 return subseqMatch (subseq, str+1);
641 static struct { const char *name; int code; }
643 { "address", TICK_ADDRESS },
644 { "unchecked_access", TICK_ACCESS },
645 { "unrestricted_access", TICK_ACCESS },
646 { "access", TICK_ACCESS },
647 { "first", TICK_FIRST },
648 { "last", TICK_LAST },
649 { "length", TICK_LENGTH },
652 { "modulus", TICK_MODULUS },
654 { "range", TICK_RANGE },
655 { "size", TICK_SIZE },
661 /* Return the syntactic code corresponding to the attribute name or
665 processAttribute (const char *str)
669 for (i = 0; attributes[i].code != -1; i += 1)
670 if (strcasecmp (str, attributes[i].name) == 0)
671 return attributes[i].code;
673 for (i = 0, k = -1; attributes[i].code != -1; i += 1)
674 if (subseqMatch (str, attributes[i].name))
679 error (_("ambiguous attribute name: `%s'"), str);
682 error (_("unrecognized attribute: `%s'"), str);
684 return attributes[k].code;
687 /* Back up lexptr by yyleng and then to the rightmost occurrence of
688 character CH, case-folded (there must be one). WARNING: since
689 lexptr points to the next input character that Flex has not yet
690 transferred to its internal buffer, the use of this function
691 depends on the assumption that Flex calls YY_INPUT only when it is
692 logically necessary to do so (thus, there is no reading ahead
693 farther than needed to identify the next token.) */
696 rewind_to_char (int ch)
698 pstate->lexptr -= yyleng;
699 while (toupper (*pstate->lexptr) != toupper (ch))
704 /* Dummy definition to suppress warnings about unused static definitions. */
705 typedef void (*dummy_function) ();
706 dummy_function ada_flex_use[] =
708 (dummy_function) yyunput